[SmileyDude]

I know there is a thread regarding panics on the general discussion forum, but I'm asking this from a developer point of view..

When a panic occurs, is the information dumped to the screen sent out somewhere? I was thinking it may be useful to either have that information written out to disk somewhere (maybe a dedicated area of the disk...) or maybe sent out as a UDP broadcast over the ethernet port... something that could be caught and then sent into Apple.

Any ideas? Is this done already?? I know you can setup debugging over ethernet... would it be fesible to have a little daemon running on another box waiting for panic broadcast messages? This could be another X/Darwin box, or maybe even a Linux, *BSD, Solaris, whatever box... just something to catch and record the panic.

------------------
dennis

[starfleetX]

That doesn't seem easy to implement. Think of the dump on your screen as a message being scrawled on a piece of paper by a man having a heart attack. The system is dying at the lowest level and these are it's dying words. And you expect it to save it to a file or broadcast it on the network? Hmm...

Sorry, I don't see that happening any time soon.

------------------
¡¡Quiero más Mac OS X!!

[SmileyDude]

Well, I'd imagine that it could write out a raw Ethernet packet(s) just as easily as it could obliterate the screen buffer with text... if not easier. I'm not asking for anything fancy here... just something that could be potentially caught.

Also, I realize that writing to disc is something that is unlikely... a floppy disc, maybe... but of course, must Macs running OS X probably don't have a floppy, and if they did, it's probably a USB one... so, it's not going there. Hard drive would open the possibility of corrupting data, not a good thing.

So, that would leave the ethernet port... something that if it was written to, isn't likely to cause problems. I would recommend some kind of packet outside of the IP protocol... this way, it wouldn't rely on having a certain IP address on the network or requiring additional configuration.

This isn't that hard to do... the fact that the kernel can catch the fatal error means that something could be done... if they wanted to, they could try to recover... but this is something that is very hard to do... especially since this is an unexpected event. In fact, with the Mach kernel, they could run the OS in a segmented off area, and when the OS panics, it could record what happened, and restart the OS inside the same kernel... of course, this would require a simplification of what the kernel does, at a cost of speed to the rest of the OS, which is why it won't be done...

but, dumping a packet (or two) out the ethernet port as it's dying would be a completely acceptable (and reasonable) alternative...

------------------
dennis